Log-Lik Kernel in CHECK PLOTS graph

Dynare help (legacy posts)
1

Hi,

Could you please explain what is the meaning of this curve plotted in the CHECK PLOTS graph?

For some graphs I get a fairly flat green line (Log-Lik Kernel) over the blue curve (log-post). By the way, the mode (sky-blue vertical line) coincides with the maximum of the blue curve. Is this ok? Or there is a problem?

Thank you very much,

V.

2

Hi,

It means that for these parameters there is no information in the data. Because the likelihood is (almost) flat, the prior distribution will show up in the posterior distribution.

This is normal. You have a problem when the vertical line (the posterior mode obtained from the optimization routine) does not coincide with the maximum of the blue curve.

Best, Stéphane.

3

Dear all,
I was wondering how can i trance the data that dynare 4.1.1 generates and uses in ploting the check plots (the log-post and the log-lik kernel)

Thanks

4

Hi, These data are not saved. You have to hack the file mode_check.m to save the values of the likelihood and the posterior kernel.
Best, Stéphane.

5

I would like to know the interpretation of the new check plots drawn by dynare 4.1.1 (what are the log-post and the log-lik kernel and what they represent, and what does the red dots that appear in some of the plots mean).

Thanks again

6

Mr Adjemain said the vertical line should coincide with the maximum of the blue one. However I keep on getting the vertical line where the green and the blue ones cross. What does that mean?

Thanks! Marcin

7

Hey hbadrakhan, I was wondering if you found out what the red plot meant please…I am having the same on some of my plots.

Thanks

8

Hi all,

What does it mean when the log posterior (blue line) is flat ? The same interpretation as when the green line is flat? (the data doens´t help to identify the parameter?). I was wondering if there is something wrong with the psi and se_rstar parameters below, particularly the latter. The identification command shows that psi is weakly identified.
Thanks in advance.
identification.pdf (5.16 KB)
mode.pdf (6.57 KB)

9

The green line tells you whether your data is informative. The change from the green to the blue line tells you what impact your prior has. The flat posterior tells you that after accounting for the data and the prior, the parameter is still weakly identified.

10

Thank you!